C# 邮递员？或者Vector.Dot那么慢？LocateFirstFoundByte方法只执行异或、乘法和移位。我怀疑向量。点会比这个快。实际上有两个重载。使用最上面的一个，它称为最下面的一个。顶部重载首先将向量视为向量，并找到第一个传递到底部重载的非零ulo

---

邮递员？或者

Vector.Dot

那么慢？LocateFirstFoundByte方法只执行异或、乘法和移位。我怀疑向量。点会比这个快。实际上有两个重载。使用最上面的一个，它称为最下面的一个。顶部重载首先将

向量

视为

向量

，并找到第一个传递到底部重载的非零

ulong

。问题在于！

    // One-time initialized vector containing { 1, 2, 3, 4, ... }
    Vector<ushort> indexes = MemoryMarshal.Cast<ushort, Vector<ushort>>(Enumerable.Range(1, Vector<ushort>.Count).Select(index => (ushort)index).ToArray())[0];

    // The input set and the element to search for
    Span<ushort> set = stackalloc ushort[]{ 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25 };
    ushort element = 22;

    // Interpret input set as a sequence of vectors (set is assumed to have length power of two for brevity)
    var setVectors = MemoryMarshal.Cast<ushort, Vector<ushort>>(set);

    // Create a vector that contains the target element in each slot
    var elementVector = new Vector<ushort>(element);

    // Loop per vector rather than per element
    foreach (var vector in setVectors)
    {
        // Get a mask that has a 1 in the single matching slot, or only 0s
        var mask = Vector.Equals(vector, elementVector);

        // Get the dot product of the mask and the indexes
        // This will multiple each index by 0, or by 1 if it is the matching one, and return their sum, i.e. the matching index or 0
        // Note that the indexes are deliberately 1-based, to distinguished from 0 (no match)
        var index = Vector.Dot(indexes, mask);

        // Either return 0 for no match, or reduce the index by 1 to get the 0-based index
        return index == 0 ? -1 : index - 1;
    }